bioRxiv preprint doi: https://doi.org/10.1101/2020.10.09.333633; this version posted October 10, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. CELL ATLAS OF THE CHICK RETINA: SINGLE CELL PROFILING IDENTIFIES 136 CELL TYPES Masahito Yamagata*, Wenjun Yan*, and Joshua R. Sanes Center for Brain Science and Department of Molecular and Cellular Biology Harvard University Cambridge MA, 02138 Running title: Chick retinal cell atlas *equal contribution For correspondence: Joshua R. Sanes Center for Brain Science Harvard University 52 Oxford St. Cambridge MA 02138 617-496-8683 [email protected] 1 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.09.333633; this version posted October 10, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. SUMMARY Retinal structure and function have been studied in many vertebrate orders, but molecular characterization has been largely confined to mammals. We used single-cell RNA sequencing (scRNA-seq) to generate a cell atlas of the chick retina. From ~40,000 single cell transcriptomes, we identified 136 cell types distributed among the six classes conserved across vertebrates – photoreceptor, horizontal, bipolar, amacrine, retinal ganglion and glial cells. To match molecular profiles to morphology, we adapted a method for CRISPR-based integration of reporters into selectively expressed genes. For Müller glia, we found that transcriptionally distinct cells were regionally localized along the anterior-posterior, dorsal-ventral and central-peripheral retinal axes. We also identified immature photoreceptor, horizontal cell and oligodendrocyte types that persist into late embryonic stages. Finally, we analyzed relationships among chick, mouse and primate retinal cell classes and types. Taken together, our results provide a foundation for anatomical, physiological, evolutionary, and developmental studies of the avian visual system. (150 words) INTRODUCTION The retina is about as complex as other regions of the vertebrate central nervous system. It differs from many regions, however, in being particularly accessible to study. For example, its neurons can be imaged live without surgical intervention, visual stimuli can be precisely controlled in time and space, and the paucity of long-distance inputs facilitates analysis of the entire circuit ex vivo. These and other technical advantages, along with the intrinsic importance of the retina and the fact that most blinding diseases arise from retinal dysfunction, have combined to make the retina a popular model for analysis of neural structure, function, development and disease (Wässle, 2004; Dowling, 2012; Hoon et al., 2014). Accordingly, retinas of many vertebrate species have been studied in detail, including those of rodents (e.g., mice and rats), carnivores (e.g., cats and ferrets), primates (e.g., macaques, marmosets and humans), birds (e.g., chickens and pigeons), fish (e.g., zebrafish and goldfish), reptiles (e.g.,turtles and lizards) and amphibia (e.g., salamanders and frogs)(Dowling, 2012; Thoreson and Dacey, 2019; Baden et al., 2020). All of these studies depend on classification and characterization of the cell types that comprise the retina. Recently, this enterprise has been greatly enhanced by the introduction of methods for high throughput single cell transcriptomic profiling (scRNA-seq), which enable comprehensive and minimally biased sampling of cell types. To date, however, these methods have been applied only to mice and primates (Macosko et al., 2015, Shekhar et al., 2016, Rheaume et al., 2018, Peng et al., 2019, Tran et al., 2019, Yan et al., 2020a,b), restricting use of non-mammalian models and making it difficult to draw evolutionary relationships among species at the molecular level . Here, we address this limitation by using scRNA-seq to generate a cell atlas of the chick retina. The basic plan of the retina is highly conserved among all vertebrates. Five neuronal classes are arranged in three cellular (nuclear) layers separated by two synaptic (plexiform) layers: photoreceptors (PRs) in an outer nuclear layer (ONL), three sets of interneurons (horizontal, bipolar and amacrine cells; HCs, BCs, ACs) in an inner nuclear layer (INL), and output neurons (retinal ganglion cells, RGCs), along with some ACs, in 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.09.333633; this version posted October 10, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. a ganglion cell layer (GCL; Cajal, 1892; Masland, 2012; Figure 1A-B). PRs form synapses with HCs and BCs in an outer plexiform layer (OPL), while RGCs, BCs and ACs form synapses in an inner plexiform layer (IPL). Axons of RGCs then exit the eye and travel through the optic nerve to a variety of retinorecipient areas in the brain (Dhande et al., 2015; Martersteck et al., 2017). Each of these classes is divided into types with specific patterns of connectivity among them endowing distinct RGC types with sensitivities to different visual stimuli, such as edges, moving or oriented objects, and color contrasts (Sanes and Masland, 2015). The retina also contains glial cells: Müller glia in the INL, and in many species, astrocytes and oligodendrocytes in and beneath the GCL (Reichenbach and Bringmann, 2013; Vecino et al., 2016). Altogether, transcriptomic and morphological studies have identified a total of >130 neural (neuronal and glial) cell types in mice and ~70 in primates (Yan et al., 2020a,b; Peng et al., 2019) Although the basic retinal plan is conserved, cell types and patterns of connectivity vary among species, serving their visual needs (Baden et al., 2020). Birds are highly visual animals with sizable eyes, generally high acuity and sophisticated retinas (Cook, 2000; Seifert et al., 2020). For example, whereas most mammals have two cone photoreceptor types, most birds are tetrachromatic, with cone photoreceptors selectively sensitive to red, green, blue, and ultraviolet light (Hart, 2001; Baden and Osorio, 2019). Many have regions specialized for high acuity vision, akin to the fovea found in primates. Indeed, histological and immunohistochemical studies have suggested that there may be more cell types in avian retinas than in those of mammals (Cajal, 1892; Mariani and Leure-duPree, 1977; Hayes, 1982; Quesada et al., 1988; Naito and Chen, 2004; Karten and Brecha, 1983; Brecha et al., 1984; Karten et al., 1990). Among birds, the retina of the domestic chicken (Gallus gallus domesticus) has been the most intensively studied. In particular, it has been a favored model for developmental analyses, including studies on the generation and migration of retinal neurons, their diversification into classes and types, the growth and guidance of RGC axons to the optic tectum, and the capacity of retinal neurons and their axons to regenerate (Adler 2000; Mey and Thanos, 2000; Thanos and Mey, 2001; Wilken and Reh, 2016; Wisely et al., 2017). To complement and facilitate these studies, we used scRNA-seq to profile cells from the chick retina. From ~40,000 single cell transcriptomes, we identified cells of all 6 classes named above (PR, HC, BC, AC, RGC and glia) and used unsupervised methods to divide them into ~150 clusters. We show that 136 of the groups represent putative cell types, with others corresponding to developmental intermediates. We then devised a method for CRISPR-based somatic cell integration of fluorescent reporters into genes shown by scRNA-seq to be expressed by specific types. Using this technique along with other histological methods, we matched molecular profiles to morphology for many neuronal types. We also found a positional signature in Müller glia, with distinct expression patterns based on their location along the anterior-posterior, dorsal-ventral and central-peripheral retinal axes. Finally, we compared the cell classes and types of chick retina with those of three mammalian species – mouse, macaque and human – demonstrating conserved molecular features of all classes and some types, along with multiple differences 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.09.333633; this version posted October 10, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. between chick and mammals. Together, our results provide new insights into retinal structure and evolution, as well as a foundation for anatomical, physiological, and developmental studies of avian retina. RESULTS Profiling chick retinal cells All known chick retinal cell types are born by E14, retinal structure is relatively mature by E18, and birds are visually competent at hatching (E21; Figure 1C; Hamburger and Hamilton, 1951; Prada et al., 1991; Cepko et al., 1996; Mey and Thanos, 2000; Yamagata and Sanes, 1995a,b; Drenhaus et al., 2003). We used a droplet-based method (Zheng et al., 2017) to obtain 30,022 high quality single cell transcriptomes from embryonic day 18 (E18) chick retina (Figure 1— figure supplement 1A). We assigned cells to classes based on expression of previously established markers, using methods described in Peng et al. (2019) and Yan et al. (2020a). We identified five neuronal classes (PRs, HCs, BCs, ACs, and RGCs) as well as two glial types, Müller glia and oligodendrocytes (Figures 1D,E). Of the E18 retinal cells profiled, 620 were RGCs (Figures 1E). This fraction (~2%) was similar to that observed in other species (Macosko et al., 2015, Peng et al., 2019) but was insufficient for extensive classification of what we anticipated would be a highly heterogeneous class.